MCP server for interacting with the Mercadona online grocery store API, enabling product search, category browsing, and shopping cart management.
The server provides 7 tools with basic descriptions and schemas, but exhibits significant quality gaps. Tool naming is mostly clear (verb-noun pattern), but descriptions lack guidance on when/why to use each tool relative to others. Parameters have type definitions and basic descriptions, but lack validation constraints, enums, and dependency hints. Output is returned as formatted strings rather than structured JSON, preventing downstream tool chaining. Error handling provides some context but lacks recovery guidance. No pagination support despite tools returning lists. The create_mercadona_cart tool is the only write operation but lacks confirmation patterns or dry-run support.
Sends a product list to the server to create a Mercadona cart. Args: products (list[int]): The list of product IDs to be sent in the cart. Returns: str: A confirmation message indicating the cart was sent successfully.
Fetches categories from the Mercadona API. Returns: str: A formatted string containing the names of the categories and the ID.
Gets the cheapest product from a subcategory (and type if specified). Args: subcategory_id (str): The ID of the subcategory to search in. type (str): The type of product to filter. Returns: str: A formatted string containing the ID and price of the cheapest product found.
Gets products by subcategory from the Mercadona API. Args: subcategory_id (str): The ID of the subcategory to fetch products for. Returns: str: A formatted string containing the products found in the subcategory.
Gets subcategories for a given category from the Mercadona API. Args: category_id (str): The ID of the category to fetch subcategories for. Returns: str: A formatted string containing the names of the subcategories.
Extracts types of products from a subcategory id. Args: subcategory_id (str): The id of a subcategory. Returns: str: A formatted string containing the types of products for the specified subcategory.
All tools return formatted strings instead of structured JSON. This prevents downstream tool chaining and forces LLMs to parse unstructured text. Example: search_product returns 'Here are the products found...' as a single string rather than {products: [{id, name, price}], total_count}.
No pagination support for list-returning tools. Tools like get_products_by_subcategory and search_product lack page, limit, and offset parameters. Results can be arbitrarily large, risking context window exhaustion.
create_mercadona_cart (destructive write operation) lacks confirmation or dry-run pattern. An LLM could accidentally add products to the user's cart without explicit confirmation, causing irreversible side effects.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 7 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 43 | - | v1 |
Searches for products in the Mercadona API. Args: query (str): The search query string. Returns: str: A formatted string containing the products found and their prices.
Parameter descriptions lack actionable constraints. Example: 'The ID of the subcategory to search in' does not state format (integer vs string), length limits, or valid ranges. get_types attempts int() conversion but does not document this requirement.
Tool descriptions do not explain when/why to use each tool relative to alternatives. Example: search_product vs get_products_by_subcategory both return products, but the distinction (free-text vs category navigation) is never stated. LLMs must guess which to call.
Error messages provide no recovery guidance. Example: get_types returns 'Invalid subcategory ID' but does not suggest calling get_categories or get_subcategories to find valid IDs. Error handling is bare-bones.
Tool names do not clearly distinguish closely related operations. 'get_more_cheap_product' is grammatically awkward and does not clearly indicate it returns a single cheapest product, not multiple affordable ones. Should be 'get_cheapest_product'.
No documentation of required dependencies or call sequence. Tools expect valid IDs from prior calls, but this dependency chain (categories → subcategories → products) is never stated. New users cannot discover the intended workflow.